May 27, 2025 | Parallel developments in wearable sensors, microfluidics, and machine learning have enabled the first-time use of a “smart bandage” in 20 human patients with chronic wounds. The latest version of the bandage, referred to as iCares, was showcased in a new study where fresh fluid from injured tissue was continually sampled to classify wound severity and healing potential.
Work on this “lab on skin” technology has been underway for more than five years now to help clinicians with the notoriously difficult job of monitoring chronic wounds, especially the non-healing variety, according to Wei Gao, Ph.D., professor of medical engineering at California Institute of Technology (Caltech) and an investigator with its Heritage Medical Research Institute. He and his colleagues from Caltech and the Keck School of Medicine of the University of Southern California cleared the latest hurdle on the long road to the clinic with an impressive demonstration of iCare’s performance (Science Translational Medicine, DOI: 10.1126/scitranslmed.adt0882).
Investigators showed that the smart bandage could detect three molecules—nitric oxide (NO, an indicator of inflammation), hydrogen peroxide (H2O2, a biomarker of infection), and oxygen (O2, reflective of a wound’s healing potential and status)—one to three days before patients experience symptoms. A machine learning algorithm they developed was also found to assess wounds on par with expert clinicians.
The in situ wound monitoring device is made of a flexible, biocompatible polymer strip that can be 3D printed at low cost. It features three microfluidic components that clear excess moisture from wounds and a nanoengineered sensor array providing real-time data about a half dozen biomarkers, says Gao. These include the three reactive species (NO, H2O2, and O2), plus pH and temperature.
The microfluidic section sucks fluid from the wound, shuttles it across the device onto the sensor array for analysis, and carries the sampled fluid away to outside the bandage, he says. The sensor includes a reusable printed circuit board to handle the signal processing and wireless data transmission to a user interface.
Gao and his team made headlines two years ago when they introduced the world’s first smart bandage in a research article that was published in Science Advances (DOI: 10.1126/sciadv.adf7388). Here, the technology was deployed in a rodent model as a wearable patch that both sensed physiological conditions of the wound bed and delivered combination therapy with antimicrobial treatment and electrically stimulated tissue regeneration to treat infected chronic wounds.
Since that time the device has received a few upgrades improving its performance, notably a means to do continuous wound fluid sampling, says Gao. “We [initially] put a passive bandage on the wound and tried to directly measure the wound fluid, but... when new fluid comes out it mixes with the old [fluid].” This would potentially change concentration levels of the wound biomarkers rather than provide real-time, accurate information.
Moreover, too much fluid isn’t good for the wound because it impairs blood flow, oxygen delivery, and nutrient transport to the wound bed. “That’s why in this [new] work we introduce nonspurious microfluidics so we can sample only the fresh fluid,” Gao says.
The iCares multiparameter sensor measurements have also been paired with a machine learning algorithm to classify wound severity and the potential for healing. That it was successfully used in 20 human patients to clinically evaluate their diabetic foot ulcer or venous leg wound—two of the most common chronic wounds—suggests it could aid in treatment decision-making.
Development of the smart bandage began with the wearable sensor to address the huge unmet need for a system to monitor the healing of chronic wounds, which represent a “huge financial burden” in the U.S. and globally, says Gao. The device was designed to act like a pump without employing one, moving fluid from one direction to another toward the sensors.
“Wound fluid monitoring is closely associated with wound fluid management, [so] we don’t want new fluid mixed with old fluid,” Gao says. “We also don’t want a wound to remain too wet,” which sets the site up for infection rather than healing. Microfluidics provided a simple answer.
Design inspiration came from nature, including plants that use capillary force to transport water from the soil through their roots and up to their leaves, says Gao. The phenomenon is related to hydrophilicity (an attraction to water) and hydrophobicity (a repulsion to water), whereby the outer surfaces of some plant tissues can become more hydrophobic, affecting water uptake.
While plants have cell membranes that control the passage of substances, the iCares device incorporates an artificial “Janus membrane” whereby fluid gets transported from the hydrophobic to the hydrophilic side, he explains. Structurally, it also has wedge-shaped channels and three-dimensional graded micropillars to help continuously transport fluid away.
For iCares to be developed, several crucial hurdles first had to be overcome, says Gao. For starters, sensors have attracted much attention over the past decade but not until recently have people paid attention to using them to monitor chemicals to gain biological insights from fluids, including saliva and sweat as well as those produced by wounds.
It also took a while for biomedical engineers to realize that attaching a sensor to a biofluid confounded the measurement process, he adds, boosting interest in using microfluidics. Once that component was added, attention turned to making the medical adhesive—the only portion of the device in direct contact with the wound—biocompatible with the human subjects who are the intended users.
Additionally, the team came to recognize the power of machine learning in turning collected sensor data into meaningful, quantitative insights for informing clinical care, including predictions about how long a wound will take to heal, says Gao. “This information, in general, is very hard to get. You need a very experienced clinician to do the evaluation and [predict the phenotype]... and even then, it is very subjective.”
While bacterial colonization is a natural occurrence with chronic wounds and can slow down the healing process, infection involves a significant bacterial burden that is the main contributor to non-healing wounds and can also be difficult to detect, he says. They don’t heal because infections keep recurring.
Stopping non-healing wounds from occurring in the first place requires treatment before infections begin, but the signs (e.g., increased pain, redness, and pus) are not always reliable or clinically detectable. “If there is a latent infection, the patient may not even realize it,” says Gao. “A sensor can give an early indication [of one]... when it is relatively easy to treat.”
Gao and his team now plan to take their research in a few different directions, including a clinical trial in a larger patient population. To that end, funding support is being sought from both the National Institutes of Health and the Defense Advanced Research Projects Agency.
In the meantime, the small clinical study continues, Gai adds. Since iCares is intended to be a disposable smart bandage destined for commercialization, another key consideration is how it will be manufactured to be scalable and low cost so people can afford to utilize it.
The overall smart bandage market is predicted to mushroom to $3.12 billion globally by the end of 2034, driven by the rising prevalence of diabetes, obesity, and cardiovascular disease. The iCares smart bandage would likely be amenable to monitoring different types of chronic wounds, says Gao, including burn wounds that require longer-term management.